A ‘History-friendly’ Way to Look at Tech Innovation
Innovative
activity is what mainly separates the winners from the losers as industries
evolve. The book Innovation
and the Evolution of Industries puts forward a new way of looking at
this central mechanism of economic growth: a systematic but ‘history-friendly’
view that takes into account the differences in industry context, as
exemplified in the computer, semiconductor and pharmaceutical industries.
In this interview with Knowledge@Wharton, Wharton
emeritus management professor and co-author Sidney Winter discusses the book
and the years-long collaboration it took with his colleagues. Winter is a Core
Team member of the Mack Institute for Innovation Management.
An edited transcript of the conversation
follows.
Knowledge@Wharton: Tell
us about your book.
Sidney Winter: Well,
I should probably first tell you about my three co-authors. I’ve got one
American co-author, my friend Richard Nelson with whom I’ve worked a lot over
the years. [They co-wrote the seminal book on evolutionary economics, An
Evolutionary Theory of Economic Change.] And then there are two Italian
colleagues, [Bocconi University professor] Franco Malerba and [IUSS professor]
Luigi Orsenigo. We’ve been at this together for quite a few years now.
We started on the research program way back in the late 1990s,
and then the first draft of the book appeared in 2012, I think it was. We
worked on it for four-and-a-half years before it was finally published by
Cambridge University Press. In terms of the subject, the book deals with the
interactions between innovation and industry evolution, how innovative activity
affects industry evolution, how industry evolution affects the level of
innovative activity. And an important part of our objective is to put forward a
new method for studying this kind of question. This new method is called
‘history-friendly modeling.’
Knowledge@Wharton: All
four of the authors are economists, I believe. Is the term ‘industry evolution’
a standard one in economics?
Winter: Yes, we are all
economists, and the answer is that, unfortunately, the term ‘industry
evolution’ is not particularly standard in economics. I suspect that most of
today’s undergraduate economics majors never hear that term in the course of
their education. But outside of economics, in innovation studies and in the
field of strategic management — also in evolutionary economics — it’s a very
familiar term.
Knowledge@Wharton: What
does evolutionary mean in the context of economics?
Winter: It means that there’s a
lot of emphasis on how economic events unfold in time, especially over
substantial periods of time. In particular, there’s the question of where new
things come from, which is the innovation part of the story. This is pretty
much what “evolutionary” means in biology, too: It’s about things unfolding
over long periods; it’s also about where the new things come from. The book
actually provides a strong illustration of these two themes, since it deals
with the interactions between evolution and innovation. And it also illustrates
a number of other aspects of the evolutionary approach.
Knowledge@Wharton: The
book goes on to examine the evolution of particular industries, does it not?
Winter: Yes, indeed it does. It
has three core chapters that deal, in turn, with the U.S. computer industry,
with the semiconductor industry — considered as a supplier to the computer
industry — and with the pharmaceutical industry. In each of these cases, we examine
a period of about 50 years or so of actual industry history. And in our
presentation, the first thing we do is to summarize that 50-year history. Then
we draw on the literature that already exists about that industry in that time
period, and consider the explanations it offers or the mechanisms it suggests
for why the industry developed in the particular way that it did.
Then, drawing on that, we create a custom-made computer
simulation model for that particular industry, in that particular time period,
and try to build into it the mechanisms that have been previously identified by
other scholars as being the important ones in history. Then we use that
simulation model, and we use it first to try to reproduce some main features of
the history itself — to show that, indeed, the explanations that have been
suggested can be made to work when they’re spelled out in detail, in the
context of a computer model.
Having done that, we take up counterfactual history. We consider
what would have happened if some of the background conditions of the industry
had been quite different. For example, suppose that the advances achieved in
semiconductor technology had been smaller than they were, or suppose that the
pharmaceutical industry had a different sort of patent system facing it than
the one it actually did. We examined those results to suggest what difference
it would have made if those particular historical circumstances had been
different. At the end of the book, we pull it all together in a summary, and we
also speculate a bit about where else one could go with these kinds of methods.
Knowledge@Wharton: Can
you give us a general idea of how these computer models work?
Winter: Yes, I certainly can.
The basic aspects of these models are the same as some that have existed for a
lot longer, in particular the same as some that [Richard] Nelson and I put
forward back in 1982, in our book An Evolutionary Theory of Economic
Change. A lot of people followed that lead and did similar work after that.
So those basic building blocks are, first of all, we have model
firms — individual firms are modeled as agents. This is sometimes called
‘agent-based modeling’ nowadays, but we were doing it before it was called
that. And then these firms are put into a model market environment where they
compete, and they set the usual things — there is price and output
determination — as they do in basic economic models. Then we also have some
sort of technological environment, or a technological opportunity environment,
which determines what it is that the firms are able to do by spending resources
on research and development activity. That research and development activity
and that particular sort of environment, drives the progress in the model and
provides feedback to profitability and other things.
We also have some other features in each of these models,
because there are always some other aspects of the setting that you have to
consider. For example, there are things that determine the conditions on which
firms exit the industry, and the conditions on which new firms enter the
industry, which are quite important. So, we have to include the specifications
of those things, too.
Knowledge@Wharton: Are
the model firms representations of actual historical companies, like IBM, Dell,
and Intel?
Winter: No, we don’t actually
attempt to match our model firms to individual historical examples. What we’re
hoping is that if we get the causal forces right — that if we understand the
mechanisms that are shaping things — then our array of toy firms, if you want
to call them that, the model firms, are going to represent the collection of
behaviors that the real firms as a collection also represented.
There’s one exception to that, or a partial exception to that,
which is kind of interesting, and it illustrates the way the model works. As
you probably know, IBM was a dominant firm of the U.S. computer industry for a
very large chunk of its history — perhaps, roughly, 35 or 40 years of its
history. And so, there is one dominant firm like that which is a salient
feature of the particular history. It’s something you ask yourself about, if
you look at that history — why did that happen, and is that related to basic
conditions of the industry?
We went at it with the assumption that it was related to the
basic conditions of the industry and that some of the causal mechanisms that
have been talked about were the relevant ones. So, we tried to build those
causes into the model. Now, when we looked at simulation results, it turned out
that, indeed, there was often one big dominant firm. We used to have the habit
of looking at output and saying, “Well, there’s IBM,” identifying this little
model firm with the historical IBM because it came to resemble it, in terms of
its role in the industry. But that wasn’t designed in. That was an emergent
feature, a thing that the causal forces in the model produced by themselves,
not something that we designed into it.
Knowledge@Wharton: When
you study the history of a particular industry, what features do you look for?
Winter: There are a number of
things that are pretty well recognized in the industry evolution literature as
being particularly important or characteristic of these patterns. And maybe the
most famous of those features is the phenomenon known as the ‘shakeout,’ which means
that typically in the start of a new industry, there is a period where there’s
a flow of new firms coming in, and some of them succeed and some of them fail,
and more come in, and more fail, and so on.
But the overall result of that is when you look at the picture,
the picture is one where, over a period of perhaps a few decades, at the start
of an industry, the number of firms involved in it goes up, up, up — and then
it reaches a peak and comes down, down, down, in quite a dramatic way. And that
‘down, down, down’ part is what’s called the ‘shakeout.’ It means a lot of
firms are failing or choosing to exit from the industry.
In the classic example of the U.S. auto industry, the number of
firms active in it peaked above 200 before the 1920s, a few decades after the
start of the industry, and then tumbled down to the “big three” over another
long period of time, as a lot of those firms fell by the wayside. That’s a very
dramatic feature of a lot of histories, and it’s one that they don’t tell you
about in the economics courses in school. It’s one of the dramatic examples of
how the evolutionary approach highlights different things than you’ve
ordinarily heard about.
Then there are some other major features. There’s the question
of what happens to the industry structure. Do a few large firms, or even one
large firm, come to dominate the scene? How does that work over a period of
time? And these processes are a reflection generally of very important feedback
loops, of success-breeds-success feedback for some of the firms involved. So,
we look for those cumulative processes and those feedback loops.
The last thing to mention is that there are sometimes
discontinuities in the technological environment, which come from sources that
are not really a feature of the behavior of the firms that are involved. For
example, in the semiconductor industry, defense-related R&D and support of
basic research relevant to semiconductors is a very, very important factor. It
shapes the opportunities, and then the relatively discrete invention of the
microprocessor is a very important discontinuity in the industry. In looking at
the history, you look for those kinds of shaping events which may not be part
of the internal logic of the economics but actually mattered to the history. And
you wouldn’t get the history back, if you left those out of the story.
Knowledge@Wharton: How
does your approach differ from what economists might do in the same area?
Winter: Well, other economists
— whom we generally refer to as ‘mainstream economists’ because they’re
following the mainstream of the research traditions that have dominated in the
discipline since about the middle of the last century — include a few who have
actually worked on the industry evolution topic, although it’s actually very few.
But in general, mainstream economics likes to focus on firms
trying to get exactly the right answers to their problems, by optimizing or
maximizing behavior, and we do not have that emphasis at all. And the reason we
don’t have that emphasis is that when you’re taking time seriously, when you’re
taking the evolutionary development seriously, you also have to take seriously
the fact that there is a lot of uncertainty in the world. And uncertainty makes
strategic decision-making by business firms a very difficult thing.
So, there are two ways to go. One way is to think harder and
harder about what firms could do to try to get exactly the right answer —
that’s the mainstream instinct. Our instinct is rather the opposite. Our
instinct is to say, “Well, mostly firms are operating out of habits or out of
rules of thumb, or out of heuristic understandings that are not precise,” and
it’s those drivers — plus luck, plus chance — that actually shape the way
behavior unfolds. We, in our emphasis on the unfolding in time, don’t put a lot
of emphasis on the effort to get exactly the right answer– that is, the firms
getting the right answer. That is a big distinction between our approach and
the mainstream approach.
Knowledge@Wharton: How
did you come up with the idea for the history-friendly approach?
Winter: Some of these concepts
and some of these methods go back quite a ways — at least they go back to our
1982 book and some of the things before that. The four of us were thinking at
one point in the 1990s about how we could do a new generation of work that was
in the spirit of that earlier work, and use some of its methods but was more
firmly dedicated to trying to understand specific pieces of reality.
This would not be the typical sort of theoretical exercise where
you showed, if this happens, that might happen. And you show that in a kind of
de-contextualized way. But rather, to try to do an exercise where we study some
actual contexts and study some actual patterns of evolution, and try to capture
in our theoretical work what it was that was going on in those particular
cases.
So that was the idea that we came across: Let’s have a new
generation that is aimed at being more directly empirically relevant, because
it is dealing with things that actually happened and trying to explain them.
That was where the idea came from.
Knowledge@Wharton:
Finally, what are some of the key takeaways from your research?
Winter: We have –- at least we
hope to reach — three different audiences. A very important audience for us is
the audience of our scientific colleagues, of our economist colleagues and the
graduate students that all of us try to raise in our own image, so to speak.
And we hope that these people will be struck by the promise of the methods that
we put forward — not just by our reading of these particular industry
histories, but rather by the fact that it is possible to be systematic in your
effort to take apart the mechanisms that were driving such a history. We hope
they will be sufficiently impressed by that, to try to do some of it
themselves, and we will have contributed to a wider movement that will improve
our understanding of the economy.
Then for managers, in particular business managers, the models
and the histories have the potential, at least, to make people think a bit
about a number of questions or strategic factors that are illustrated in these
histories. One of the first questions is, in your situation, where are you
really, relative to the shakeout? Given this very strong, typical pattern, if
you’re in an industry, and the industry is 10 to 15 years old, you may well be
on the way to the shakeout, not past it. So that’s an important thing to
understand: There’s going to be a time of stress, in which a lot of the
contenders are going to fall by the wayside.
To see that — particularly in our pharmaceutical industry
chapter, our discussion of why that happens and why it doesn’t happen some of
the time, because it doesn’t always happen — I think that would be a useful
stimulus to the strategic imaginations of managers who can match the elements
of the story to their own situation and ask where they stand in those different
respects.
And then in some cases, the results of the work, I think, have
also got an audience in the public policy sphere, where in particular some of
our work with the counterfactual history and the role of the patent system in
pharmaceuticals — some of that certainly opened my eyes about possibilities for
causal mechanisms that were rather different than I had imagined, situations
more complicated than I had imagined. And we accidentally got some of this
complication more or less well represented in the models. I mean not really
accidentally, but in making an attempt, we happened to have had the luck to
make a good enough attempt to actually get some insight about the way those
things work. So that’s another possible audience.
And then lastly, there are people who are just interested in the
way the economy works in an economic history, and the way industries work.
There’s a lot of material in the book for them. They wouldn’t necessarily have
to follow in detail our computer modeling choices that we argue for, but
there’s a lot of history, a lot of suggestive quantitative results, a lot of
interesting charts, and so on. I think that people who are just generally
interested in the economy would find a lot to appreciate there.
http://knowledge.wharton.upenn.edu/article/a-history-friendly-way-of-looking-at-tech-innovation/?utm_source=kw_newsletter&utm_medium=email&utm_campaign=2017-05-16
No comments:
Post a Comment